johnx125

Honorable
Oct 10, 2017
117
0
10,690
I have a Corsair H60 (2013 version) cooler in my computer and the CPU is an AMD FX8350 running at stock frequency. It's a pretty old rig I built it back in early 2015 and the CPU never overheated so I didn't really worry about temps. However, a few days ago I noticed that both of the cooler's tubes get hot when the CPU is under load. One tube is slightly hotter than the other but I think there should be a bigger difference. Shouldn't one tube be almost cold and the other one very hot?

P.S.: CPU temps are still fine it never exceeds 60°C under load.
 
Solution
I have a Cryorig A80 liquid cooler
2x USB powered digital temp probes.

Around the coolant pipes from the pump to the rad, there is automotive wire wrap, just for the industrial look.
The temp probes are siamesed inside the wire wrap, next to the actual coolant tubes, so they are reading the temperature of the pipe, not the actual coolant.
But this is an indicator of the possible delta between supply and return.

As I sit here and the system mostly at idle....the two temp probes read 26.5C (supply) and 26.4C (return from the rad).
Current CPU temp of 36C.

Just to be clear...that 26.5 and 26.6 temp is NOT the actual coolant temp, but rather the temp of the tubes.
But they are basically identical.

Under full load, there might be a 2...
May 4, 2019
59
6
45
If there is a difference, there is probably flow, and the fact that you have reasonable CPU temps backs that up.

I've actually measured the temperatures of my two tubes on a custom loop, and the difference is 0.5 C between the inlet and the outlet. It is rejecting up to 200 watts at that time.

If the flow rate is high, then there is little temperature delta between the inlet and outlet. The physics make that obvious:
delta T is proportional to the heat absorbed divided by the liquid volume. High flow rates mean a large volume per unit time.
 

johnx125

Honorable
Oct 10, 2017
117
0
10,690
If there is a difference, there is probably flow, and the fact that you have reasonable CPU temps backs that up.

I've actually measured the temperatures of my two tubes on a custom loop, and the difference is 0.5 C between the inlet and the outlet. It is rejecting up to 200 watts at that time.

If the flow rate is high, then there is little temperature delta between the inlet and outlet. The physics make that obvious:
delta T is proportional to the heat absorbed divided by the liquid volume. High flow rates mean a large volume per unit time.
So basically if the coolant flows really fast through the radiator, then the radiator does have enough time to fully cool down the liquid.
 

johnx125

Honorable
Oct 10, 2017
117
0
10,690
FYI the max temp for that CPU is 61C, so you may be on the verge of overheating. Although many utilities don't measure FX CPU temps properly so the most accurate way to tell is to look at thermal margin in AMD Overdrive (higher is better, if it hits 0 you're overheating).
Well it's not always at 60C... usually under load it runs at 55C and only spikes up to 60 for 0.5 seconds. The highest temperature reading I've seen with this cooler is 64C during a stress test that lasted 2 hours (very stupid, I know :p). Also it's pretty hot in my country during the summer and ambient temperature in my house is always around 28C this time of year.
 

Karadjgne

Titan
Ambassador
Cpu temp is not coolant temp. Liquids have a huge ability to absorb energy, in this case heat, without it affecting the coolant much, if any.

Take a pan and put a good amount of water in it. Put the burner on low, (that's @ 300w) and see how long it takes to get hot.

What the radiator does is dissipate the energy transfered to the coolant. The coolant itself is only @ ambient temp + @ 10°C ish , upto 20°C ish under heavy loads. So in your case, it's running in the low 40's to 50ish.

Fx Cpus do not have a thermal sensor in the cores, so any program trying to read those like they'd read Intel are going to be wrong, instead reading other sensors around the socket that will not be accurate for the cpu. There's only 2 programs that will accurately read FX cpus, and thats AMD Overdrive and Coretemp, set for Thermal Margin. This is where the program reads the cpu TjMax (absolute max safe temp) and through extensive algorithms using data from various sources, comes up with a number representing how close you are to TjMax. The actual number itself is unimportant, what the number represents is everything. If your TM is in the 30's you've got plenty of room left for heat, no worries. If TM is in 20's, alls still good. If TM is 20-10, now you are warm, might want to watch it. 10-0 you are definitely running hot, time to fix that with lowering OC, better cooler, mess with fan curves, change settings etc. If you ever see a TM with a negative number, the cpu is cooking, needs immediate fix.

In the past, many adopted Package temp as their temp of choice, but applied it to Intel standards. Package temps are usually @ 10°C hotter than the figured cpu temp. Which gave them a 70°C safe zone, even though actual rated core temp limit is 62°C for all the FX.

Best bet is as TJ Hooker suggested, run Overdrive and check your thermal margin. The range it's in will be a far better guide than the questionable results you are currently seeing.
 
  • Like
Reactions: TJ Hooker

USAFRet

Titan
Moderator
I have a Cryorig A80 liquid cooler
2x USB powered digital temp probes.

Around the coolant pipes from the pump to the rad, there is automotive wire wrap, just for the industrial look.
The temp probes are siamesed inside the wire wrap, next to the actual coolant tubes, so they are reading the temperature of the pipe, not the actual coolant.
But this is an indicator of the possible delta between supply and return.

As I sit here and the system mostly at idle....the two temp probes read 26.5C (supply) and 26.4C (return from the rad).
Current CPU temp of 36C.

Just to be clear...that 26.5 and 26.6 temp is NOT the actual coolant temp, but rather the temp of the tubes.
But they are basically identical.

Under full load, there might be a 2 or 3 C difference between supply and return. Nothing huge.

A well functioning radiator will have not a lot of delta between supply and return.
 
  • Like
Reactions: TJ Hooker
Solution

johnx125

Honorable
Oct 10, 2017
117
0
10,690
Cpu temp is not coolant temp. Liquids have a huge ability to absorb energy, in this case heat, without it affecting the coolant much, if any.

Take a pan and put a good amount of water in it. Put the burner on low, (that's @ 300w) and see how long it takes to get hot.

What the radiator does is dissipate the energy transfered to the coolant. The coolant itself is only @ ambient temp + @ 10°C ish , upto 20°C ish under heavy loads. So in your case, it's running in the low 40's to 50ish.

Fx Cpus do not have a thermal sensor in the cores, so any program trying to read those like they'd read Intel are going to be wrong, instead reading other sensors around the socket that will not be accurate for the cpu. There's only 2 programs that will accurately read FX cpus, and thats AMD Overdrive and Coretemp, set for Thermal Margin. This is where the program reads the cpu TjMax (absolute max safe temp) and through extensive algorithms using data from various sources, comes up with a number representing how close you are to TjMax. The actual number itself is unimportant, what the number represents is everything. If your TM is in the 30's you've got plenty of room left for heat, no worries. If TM is in 20's, alls still good. If TM is 20-10, now you are warm, might want to watch it. 10-0 you are definitely running hot, time to fix that with lowering OC, better cooler, mess with fan curves, change settings etc. If you ever see a TM with a negative number, the cpu is cooking, needs immediate fix.

In the past, many adopted Package temp as their temp of choice, but applied it to Intel standards. Package temps are usually @ 10°C hotter than the figured cpu temp. Which gave them a 70°C safe zone, even though actual rated core temp limit is 62°C for all the FX.

Best bet is as TJ Hooker suggested, run Overdrive and check your thermal margin. The range it's in will be a far better guide than the questionable results you are currently seeing.
I've used AMD Overdrive in the past and thermal margin is always around 50C at idle and under load it's usually around 23C maybe 20C sometimes. So it's fine if the temperature of the coolant is almost the same on both the supply and the return hoses right?
 

johnx125

Honorable
Oct 10, 2017
117
0
10,690
I have a Cryorig A80 liquid cooler
2x USB powered digital temp probes.

Around the coolant pipes from the pump to the rad, there is automotive wire wrap, just for the industrial look.
The temp probes are siamesed inside the wire wrap, next to the actual coolant tubes, so they are reading the temperature of the pipe, not the actual coolant.
But this is an indicator of the possible delta between supply and return.

As I sit here and the system mostly at idle....the two temp probes read 26.5C (supply) and 26.4C (return from the rad).
Current CPU temp of 36C.

Just to be clear...that 26.5 and 26.6 temp is NOT the actual coolant temp, but rather the temp of the tubes.
But they are basically identical.

Under full load, there might be a 2 or 3 C difference between supply and return. Nothing huge.

A well functioning radiator will have not a lot of delta between supply and return.
Ook I understand how it works now... Thank you for replying :D
 

USAFRet

Titan
Moderator
Right. The fluid is in constant motion through the system.

A large temp delta might indicate a pump that is too small or too slow, not moving enough liquid through the system. Letting the supply get too warm before it gets to the rad.

Running Prime95 v26.6 for 30 mins to let the temp of the tubes stabilize, ambient of 22C:
Starting:
Supply - 26.5
Return - 26.5
CPU - 33

After 30 mins:
Supply - 29.2
Return - 28.5
CPU - 56
 

johnx125

Honorable
Oct 10, 2017
117
0
10,690
Right. The fluid is in constant motion through the system.

A large temp delta might indicate a pump that is too small or too slow, not moving enough liquid through the system. Letting the supply get too warm before it gets to the rad.

Running Prime95 v26.6 for 30 mins to let the temp of the tubes stabilize, ambient of 22C:
Starting:
Supply - 26.5
Return - 26.5
CPU - 33

After 30 mins:
Supply - 29.2
Return - 28.5
CPU - 56
Well turns out I have a perfectly functioning pump and radiator :p. I ran a stress test last month using OCCT. I let it ran for 45 minutes and the average CPU temperature was 57°C (according to AMD Overdrive readings) and when I touched the tubes both of them were hot.
 

johnx125

Honorable
Oct 10, 2017
117
0
10,690
Stress test, under cpu limits. I'd say it's working as it should. Gaming usually only stresses a pc to @ 70% of what occt can do to a pc, so you should be well within limits.
Yup, temperatures when gaming are lower than stress testing (usually around 52 to 55°C). Also unlike stress testing, the radiator fan doesn't run at full speed when gaming so temperatures could be even lower. Btw is OCCT fine for this kind of testing or should I try another programm?
 

Karadjgne

Titan
Ambassador
Normally ppl recommend Prime95 v26.6 small fft for stress testing. It gives a clean constant 100% load that uses the same instruction sets as games, so simulates a worst case gaming scenario. Aida64, occt, IBT and other testers bounce anywhere from 80% to 130% cpu loads, so don't exactly give a clean baseline from which to work with. Which is ok too, with air cooling as that only requires @ 5-10 minute test. Not so good for liquid cooling that needs 30 minutes+ to equalize coolant temps.
 

johnx125

Honorable
Oct 10, 2017
117
0
10,690
Normally ppl recommend Prime95 v26.6 small fft for stress testing. It gives a clean constant 100% load that uses the same instruction sets as games, so simulates a worst case gaming scenario. Aida64, occt, IBT and other testers bounce anywhere from 80% to 130% cpu loads, so don't exactly give a clean baseline from which to work with. Which is ok too, with air cooling as that only requires @ 5-10 minute test. Not so good for liquid cooling that needs 30 minutes+ to equalize coolant temps.
Ok got it next time I run a stress test for temperatures I'll try Prime95. Thank's a lot for the help!! :D